dsh-geometric-mean
Run the following command in DeepSeek Harness:
dsh plugin install uckkk/dsh-geometric-mean
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install uckkk/dsh-geometric-mean in the DeepSeek Harness terminal to install this plugin; the full source is available at https://github.com/uckkk/dsh-geometric-mean
About this plugin
When you benchmark model inference, you often need to fold several metrics—perplexity, time-to-first-token, throughput—into a single composite score. Arithmetic means are easily distorted by a single outlier, while the geometric mean naturally penalises any weak dimension, making it the more reliable aggregator for inference-quality metrics.
dsh-geometric-mean is a tiny, pure Node.js library that computes the geometric mean of a numeric array. It brings zero external dependencies and makes zero network calls, so you can drop it into an inference pipeline, an evaluation script, or a CI job without worrying about supply-chain bloat or connectivity. The function is deterministic: same input, same output, every time.
If you build LLM inference benchmarks, write automated scoring scripts, or simply need an offline, reproducible way to aggregate multi-dimensional metrics without pulling in a heavyweight math library, this plugin gives you a one-import, one-call solution with no configuration overhead.
Use Cases
- Aggregating latency, throughput, and accuracy into a single composite inference score
- Calling a scoring function inside an offline CI pipeline with zero network requests
- Replacing arithmetic mean in evaluation scripts to resist outlier distortion
Best For
- Engineers building LLM inference benchmarks
- Data engineers writing automated model-scoring scripts
- Researchers aggregating multi-dimensional metrics in offline environments
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.